)]}' { "log": [ { "commit": "a15e16d11a97981f4bee6db6f11de5440169e955", "tree": "a2b5921a89a3c5bfe9986877822e09b04f1a5c23", "parents": [ "386cb629c0d003b099d22eb7c242aefcadf6f3e7" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Fri May 29 10:50:52 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 29 01:55:43 2026 -0700" }, "message": "[Flang] Move builtin .mod generation into runtimes (Reapply #137828) (#171515)\n\nReapplication of #137828 and #169638\n\nMove building the .mod files from openmp/flang to openmp/flang-rt using\na shared mechanism. Motivations to do so are:\n\n1. Most modules are target-dependent and need to be re-compiled for each\ntarget separately, which is something the LLVM_ENABLE_RUNTIMES system\nalready does. Prime example is `iso_c_binding.mod` which encodes the\ntarget\u0027s ABI. Constants such as `c_long_double` also have different\nvalues).\nMost other modules have `#ifdef`-enclosed code as well. For instance\nthis caused offload targets nvptx64-nvidia-cuda/amdgpu-amd-amdhsa to use\nthe modules files compiled for the host which may contrain uses of the\ntypes REAL(10) or REAL(16) not available for nvptx/amdgpu.\n\n2. CMake has support for Fortran that we should use. Among other things,\nit automatically determines module dependencies so there is no need to\nhardcode them in the CMakeLists.txt.\n\nCurrently the dependencies between builtin modules and `omp_lib.mod`\nseem to be incorrect:\nhttps://github.com/llvm/llvm-project/pull/174474#issuecomment-3880189581\n\n3. It allows using Fortran itself to implement Flang-RT. Currently, only\n`iso_fortran_env_impl.f90` emits object files that are needed by Fortran\napplications (#89403). The workaround of #95388 could be reverted (PR\n#169525).\n\nIf using Flang for cross-compilation or target-offloading, flang-rt must\nnow be compiled for each target not only for the library, but also to\nget the target-specific module files. For instance in a bootstrapping\nruntime build, this can be done by adding:\n`-DLLVM_RUNTIME_TARGETS\u003ddefault;nvptx64-nvidia-cuda;amdgpu-amd-amdhsa`.\n\nSome new dependencies come into play:\n* openmp depends on flang-rt for building `lib_omp.mod` and\n`lib_omp_kinds.mod`. Currently, if flang-rt is not found then the\nmodules are not built.\n* check-flang depends on flang-rt: If not found, the majority of tests\nare disabled. If not building in a bootstrpping build, the location of\nthe module files can be pointed to using\n`-DFLANG_INTRINSIC_MODULES_DIR\u003d\u003cpath\u003e`, e.g. in a flang-standalone\nbuild. Alternatively, the test needing any of the intrinsic modules\ncould be marked with `REQUIRES: flangrt-modules`.\n* check-flang depends on openmp: Not a change; tests requiring\n`lib_omp.mod` and `lib_omp_kinds.mod` those are already marked with\n`openmp_runtime`.\n\nAs intrinsic are now specific to the target, their location is moved\nfrom `include/flang` to `\u003cresource-dir\u003e/finclude/flang/\u003ctriple\u003e`. The\nmechnism to compute the location have been moved from flang-rt\n(previously used to compute the location of `libflang_rt.*.a`) to common\nlocations in `cmake/GetToolchainDirs.cmake` and\n`runtimes/CMakeLists.txt` so they can be used by both, openmp and\nflang-rt. Potentially the mechnism could also be shared by other\nlibraries such as compiler-rt.\n\n`finclude` was chosen because `gfortran` uses it as well and avoids\nmisuse such as `#include \u003cflang/iso_c_binding.mod\u003e`. The search location\nis now determined by `ToolChain` in the driver, instead of by the\nfrontend. Another subdirectory `flang` avoids accidental inclusion of\ngfortran-modules which due to compression would result in\nuser-unfriendly errors. Now the driver adds `-fintrinsic-module-path`\nfor that location to the frontend call (Just like gfortran does).\nSince the driver determines the location,\ntests invoking `flang -fc1` and `bbc` must also be passed the location\nby llvm-lit. This works like llvm-lit does for finding the include dirs\nfor Clang using `-print-file-name\u003d...`.\n\nGitOrigin-RevId: bafde6fbb860e9b2f887a649c19f3b3ebb168041\n" }, { "commit": "386cb629c0d003b099d22eb7c242aefcadf6f3e7", "tree": "9cfdab93e372443057b6321083cfbcd7961cc1dc", "parents": [ "c364fb6969ce4e9ee9b0490093469fce57d84e67" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Mon May 18 10:08:44 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon May 18 08:12:04 2026 -0700" }, "message": "[libc] Demote compiler check error to a warning (#198033)\n\nSummary:\nThis check exists to encode the policy that this is only intended to be\nbuilt with a just-built compiler. In practice it\u0027s a little too strict\nand breaks pretty much every six months when the version bumps or when\npeople try to build a separate patch. Just demote to a warning.\nGitOrigin-RevId: 13da33e922fe43cd97246f5e33320acc4f5ea186\n" }, { "commit": "c364fb6969ce4e9ee9b0490093469fce57d84e67", "tree": "8d1747ad1cea9848ee5dbaf0c2340a3baed3d46f", "parents": [ "ef851dea0499e9fe2fdbc4612e35beaad3afa7fb" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Mon May 18 08:26:00 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon May 18 06:30:39 2026 -0700" }, "message": "[Offload] Add a proper `install-offload` target (#198206)\n\nSummary:\nThis target is missing, it\u0027s the expected behavior for all LLVM\ncomponents to have a way to install them individually as components.\nThis allows us to selectively install offload versus other pieces of the\nruntime. Also make this and the openmp handling canonical with libcxx.\nGitOrigin-RevId: 499d0bb35f1c9fa0a6421c1a333de3683ff490cb\n" }, { "commit": "ef851dea0499e9fe2fdbc4612e35beaad3afa7fb", "tree": "25f7ffda0572bc783ea1f25e948f58ab97b5258b", "parents": [ "1d3f8cc725925a9755e5d604845d36efd6fe5f8e" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Fri May 15 07:37:18 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 15 05:41:42 2026 -0700" }, "message": "[OpenMP] Fix missing install-openmp component (#197603)\n\nSummary:\nThis pattern is consistent throughout all the runtimes and is what the\ntop-level `install-openmp-\u003ctriple\u003e` corresponds to. It should be\nprovided and used.\nGitOrigin-RevId: eec28baa5f6e8a368c113cc9b1a1231668217d12\n" }, { "commit": "1d3f8cc725925a9755e5d604845d36efd6fe5f8e", "tree": "c488f1ea13e08befec6b6ff33c450374c30c2ea1", "parents": [ "81265d0f4aefa2a3cc5cb358b1bd7e8982b8845b" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Wed May 13 09:26:48 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 13 00:31:17 2026 -0700" }, "message": "[OpenMP][offload] Inline target reductions (#196061)\n\nSignificantly reduces register usage and removes register spilling in\n`offload/test/offloading/multiple-reductions.cpp`, for example. Provides\nspeedup of up to 5-10x for a lot of reductions in such a larger setup.\n\nBased on https://github.com/llvm/llvm-project/pull/195940.\nSee also the discussion in\nhttps://github.com/llvm/llvm-project/pull/195102.\n\nGitOrigin-RevId: f59aca9103391db35e6f1fc06cd441701dd4f876\n" }, { "commit": "81265d0f4aefa2a3cc5cb358b1bd7e8982b8845b", "tree": "fdbd4ae76092009a6f070543f63bdfb72e207966", "parents": [ "c7f592c6eea52ce133ace5db8ab25d4d41ed4000" ], "author": { "name": "Jan Patrick Lehr", "email": "JanPatrick.Lehr@amd.com", "time": "Fri May 08 13:49:27 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 08 04:51:15 2026 -0700" }, "message": "[OpenMP] Fix set-but-unused-var warning in omptest (#196069)\n\nThis fixes a warning in omptest about a set but unused variable. The var\nwas intended to control whether colored logging output is created.\nThat logic has been moved into the `Logger` itself.\n\nGitOrigin-RevId: 27c5421df21b05def2a297028bfccc9a6ba16396\n" }, { "commit": "c7f592c6eea52ce133ace5db8ab25d4d41ed4000", "tree": "fe6d889012a46362d45cb41ab804190d4297a830", "parents": [ "ecc35813f6dc5c1fa2bd4d72919b3dd74ce6fa6a" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed May 06 16:51:14 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 06 14:55:15 2026 -0700" }, "message": "[OpenMP] Fix incorrect use of align instead of round (#196191)\n\nGitOrigin-RevId: 1232bc01626955f7b47c3b17fc8bca0f7ab9f223\n" }, { "commit": "ecc35813f6dc5c1fa2bd4d72919b3dd74ce6fa6a", "tree": "d2feb148d75495314dec0b59a385a972bd70cf7a", "parents": [ "06184ead2ba224125fd9bdf096c280432dc25708" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed May 06 15:37:43 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 06 13:40:42 2026 -0700" }, "message": "[OpenMP][NFC] Simplify rounding operations (#196155)\n\nSummary:\nThere were a lot of these cases that did rounding up / down. Make\nhelpers for them and simplify.\nGitOrigin-RevId: 4bae755a3f8a9deff95162288efc9d33795fe42f\n" }, { "commit": "06184ead2ba224125fd9bdf096c280432dc25708", "tree": "e8bab7fdc6c9c371f927d1328d19ac15ba46a329", "parents": [ "2785d7f1c018c53cb0dd9add3c62ee215b3a5187" ], "author": { "name": "Daniel Chen", "email": "cdchen@ca.ibm.com", "time": "Wed May 06 11:58:00 2026 -0400" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 06 09:00:40 2026 -0700" }, "message": "[openmp][flang] To add a cmake option to build OPENMP Fortran modules only without building the libomp lib (#195576)\n\nPR #192687 Introduces a dependency to the Fortran intrinsic module when\nbuilding `openmp` Fortran modules, which requires a Flang enabled llvm\nbuild in order to build `openmp` Fortran modules.\n\nThis PR is to add a cmake option `LIBOMP_FORTRAN_MODULES_ONLY` to allow\nusers to build \u0027openmp` Fortran modules only for Flang without building\nthe libomp runtime library.\nThis will allow libomp.a/so build still be independent from the Flang\nbuild as before.\n\nE.g. To build `openmp` Fortran modules only for Flang, one would need to\nuse the following options.\n```\ncmake \\\n...\n -DLIBOMP_FORTRAN_MODULES_ONLY\u003dON \\\n -DLLVM_ENABLE_PROJECTS\u003d\"flang\" \\\n -DLLVM_ENABLE_RUNTIMES\u003d\"openmp\" \\\n ...\n\nGitOrigin-RevId: dd145eb8997878143f648a7601741f6409330963\n" }, { "commit": "2785d7f1c018c53cb0dd9add3c62ee215b3a5187", "tree": "cf10fe7c772767f205f3ea49ec7af98dafbcc93f", "parents": [ "a0914e2014556e1fa6f5554179d3da9979ffc2d2" ], "author": { "name": "Jan André Reuter", "email": "j.reuter@fz-juelich.de", "time": "Tue May 05 16:03:58 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 05 07:05:48 2026 -0700" }, "message": "[OpenMP][OMPT] Update tools header for OpenMP v6.0 (#195829)\n\nAdd new enum values, structs and variables to `omp-tools.h` header,\naccording to the OpenMP v6.0 specification.\nUpdate `openmp` and `offload` tools headers to recognize the new values\nto ease testing once those features are implemented.\n\nSince no feature is implemented here, update the docs to mention\n`partial` support, pointing to this PR for the enums.\n\n---------\n\nSigned-off-by: Jan André Reuter \u003cj.reuter@fz-juelich.de\u003e\nGitOrigin-RevId: 8be29edc20dfb19245ab8cf49669c80125ece4f5\n" }, { "commit": "a0914e2014556e1fa6f5554179d3da9979ffc2d2", "tree": "9d74097e34aa2e5f4f0bc752cddc17de48dec1d9", "parents": [ "4739ca0f1b0c6e2d57361253651b41772475268e" ], "author": { "name": "Jan Patrick Lehr", "email": "JanPatrick.Lehr@amd.com", "time": "Tue May 05 11:40:06 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 05 02:45:55 2026 -0700" }, "message": "[OpenMP] Add target to build OpenMP runtime (#189557)\n\nThis adds a top-level target to build the OpenMP runtime, similar to\nwhat was done in https://github.com/llvm/llvm-project/pull/186099 for\nthe Offload runtime.\n\nHaving this top-level target enables us to execute the build in the\npre-commit CI as shown in\nhttps://github.com/llvm/llvm-project/pull/174955 (I actually just\ncherry-picked the commit from that branch)\n\nGitOrigin-RevId: 6d8d7d00a6de1eeb1e7cffe1ad2f5ee85b53ed9a\n" }, { "commit": "4739ca0f1b0c6e2d57361253651b41772475268e", "tree": "bd56e5a7bbcd617e799f2dab5306e1139a17d83a", "parents": [ "02c45202bfa6157050fe2872b45071e76f26006c" ], "author": { "name": "Xing Xue", "email": "xingxue@outlook.com", "time": "Thu Apr 30 16:30:30 2026 -0400" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 30 13:37:06 2026 -0700" }, "message": "[libomp] Set arch string for PPC (#195129)\n\nAdd the missing arch string for PPC.\n\nGitOrigin-RevId: f56e69527294e580e7e554996f7c7fa5318b176e\n" }, { "commit": "02c45202bfa6157050fe2872b45071e76f26006c", "tree": "e8e28ad892d272f6f3dbd3c128f73421ac511c37", "parents": [ "69e16976c92e77c60d0ed6a7bea7413c2b400475" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed Apr 29 13:10:38 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Apr 29 11:12:46 2026 -0700" }, "message": "[Offload] Clean up and split Shared/Utils.h (#194876)\n\nSummary:\nThis is used in the GPU portion of the build, but really shouldn\u0027t be.\nMost of these helpers were only used by the GPU portion, which has much\nnicer builtins available in new clang. Using these in the main offload\nbuild is broken on Windows and not available on older compilers, so we\nsplit. Also means one less header we share.\n\nThe alignment helpers are trivially replaced by an LLVM utility, the\nonly thing that remains are the pointer arithmetic functions. There\u0027s no\nconveient place to put these so I just let them stay.\n\nGitOrigin-RevId: 105bd6b654ddf4c42626dfed6efd4e99a9a1dddf\n" }, { "commit": "69e16976c92e77c60d0ed6a7bea7413c2b400475", "tree": "94d5bd5cc66e612e05c825486c725807097ec4f1", "parents": [ "f1698710204d0e417be8d668636d310456042be8" ], "author": { "name": "Jan André Reuter", "email": "j.reuter@fz-juelich.de", "time": "Wed Apr 29 12:06:41 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Apr 29 03:10:23 2026 -0700" }, "message": "[OMPT][OpenMP] Use omp_initial_device for host in callbacks (#192924)\n\nThe OpenMP specification offers different ways for identifying the host\ndevice. While users of the OpenMP API can use `omp_get_initial_device()`\nor the constant `omp_initial_device` (available since OpenMP v5.2), a\ntool needs to rely on the `initial_device_num` passed by the OpenMP\nruntime during the `initialize` callback.\n\nIn #134451, it was discovered that the `initial_device_num` passed is\nalways `0`, regardless of any device are available for offload\nexecution. For host-only OpenMP code, this matches the result of\n`omp_get_num_devices()`, and is a valid result. In the case of devices\nbeing available though, this passed identifier is incorrect. While\n`libomp` calls `omp_get_num_devices()`, `libomptarget` has not fully\ninitialized its PluginManager at that point, hence returning no\navailable devices. Tools relying on `initial_device_num` might therefore\nincorrectly assume host-side execution when some code runs on a device.\nSince the `ompt_get_num_devices()` entry point is also not fully\nimplemented, tools currently need to do on-the-fly handling for the host\ndevice.\n\nTo make handling easier, consistently use the `omp_initial_device`\nidentifier as the `initial_device_num` and for all device callbacks.\nThis does not rely on knowing the number of available devices during\nOMPT initialization in the runtime and ensures a consistent value for\ntools to handle.\n\nCloses #134451\n\nSigned-off-by: Jan André Reuter \u003cj.reuter@fz-juelich.de\u003e\nGitOrigin-RevId: 0a4798ad4b9612f9499f4c13104603a4c94a6144\n" }, { "commit": "f1698710204d0e417be8d668636d310456042be8", "tree": "e8d02dc0df5032fa66b18e3c6cfbaa3ed91b83fd", "parents": [ "df6892583ab629dd1cd96d23b268b898238034b6" ], "author": { "name": "Mark Zhuang", "email": "mark.zhuang@spacemit.com", "time": "Wed Apr 29 05:05:33 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 28 14:10:44 2026 -0700" }, "message": "[OpenMP] Fix use of uninitialized Dl_info when dladdr fails in ompd_init (#193643)\n\nWhen dladdr() fails, dl_info is left uninitialized. The code printed the\nerror but then fell through to unconditionally call\nstrrchr(dl_info.dli_fname, \u0027/\u0027), accessing an uninitialized pointer --\ntypically NULL on the stack -- causing a SIGSEGV.\n\nFix by turning the second \u0027if\u0027 into \u0027else if\u0027, so the path-extraction\nblock is only reached when dladdr() succeeded and dl_info is valid.\n\nAssisted-by: Claude Sonnet 4.6\nGitOrigin-RevId: ab43ab6158a19f60048cdbc77fe645ae61de0bae\n" }, { "commit": "df6892583ab629dd1cd96d23b268b898238034b6", "tree": "6d8db8b57433f4c75a38d53d63cfd1f801dffc07", "parents": [ "92396f22e2432f71c71885ce0788f7287ad48801" ], "author": { "name": "Amit Tiwari", "email": "amtiwari@amd.com", "time": "Tue Apr 28 21:31:04 2026 +0530" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 28 09:05:15 2026 -0700" }, "message": "[Flang][OpenMP] Validate `omp_initial_device` `omp_invalid_device` as device IDs (#193669)\n\nAs per OpenMP 5.2/6.0 the below are valid device values in a `#pragma\nomp target` directive:\n\nomp_initial_device (-1) -\u003e refers to the host CPU.\nomp_invalid_device (-2) -\u003e an intentionally invalid device, used to\ntrigger a runtime error.\n\nFor the 2 values discussed above flang fails with:\n\n```\nerror: The device expression of the DEVICE clause must be a positive integer expression\n !$OMP TARGET DEVICE(-1)\nerror: Must have INTEGER type, but is REAL(4)\n !$OMP TARGET DEVICE(OMP_INVALID_DEVICE)\n\n```\nIssue: https://github.com/llvm/llvm-project/issues/192989\nGitOrigin-RevId: 3c14034c55a296306ad0ea4990f0f1b34e9e5d6e\n" }, { "commit": "92396f22e2432f71c71885ce0788f7287ad48801", "tree": "cb39f8de5eac60a7f495c53d9dc3936c261da023", "parents": [ "a4722521ebecd258dd7d8d7d96c4a6c10ff14284" ], "author": { "name": "Julian Brown", "email": "julian.brown@amd.com", "time": "Tue Apr 28 10:38:09 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 28 02:41:14 2026 -0700" }, "message": "[OpenMP] Rename ompx_taskgraph-\u003eomp_taskgraph_experimental\n\nThis patch renames the option to enable taskgraph support in the\nruntime from OMPX_TASKGRAPH to OMP_TASKGRAPH_EXPERIMENTAL, to reflect\nthe feature\u0027s official status in OpenMP 6.0, but also the feature\u0027s\ncurrent work-in-progress nature.\n\ncommit-id:fa62775a\n\nReviewers: ro-i\n\nReviewed By: ro-i\n\nPull Request: https://github.com/llvm/llvm-project/pull/194045\n\nGitOrigin-RevId: cced4082785cde51cc76d2df76066a5a3feff232\n" }, { "commit": "a4722521ebecd258dd7d8d7d96c4a6c10ff14284", "tree": "df237b6dc0ddb7f3c8b4c623fef221154fcb53de", "parents": [ "577075d233e61808edf5bb2a3ac149dc62ba2529" ], "author": { "name": "Amit Tiwari", "email": "amtiwari@amd.com", "time": "Tue Apr 28 14:28:37 2026 +0530" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 28 02:00:28 2026 -0700" }, "message": "[Clang][OpenMP] Validate omp_initial_device omp_invalid_device as device IDs (#193688)\n\nThe counterpart fix for clang (as too done here:\n[flang-fix](https://github.com/llvm/llvm-project/pull/193669))\n\nThe incorrectly interpreted device values in the `target` directive\nthrows:\n\n```\n error: argument to \u0027device\u0027 clause must be a non-negative integer value\n #pragma omp target device(-1)\n ^~\nerror: argument to \u0027device\u0027 clause must be a non-negative integer value\n #pragma omp target device(omp_invalid_device)\n ^~~~~~~~~~~~~~~~~~\n```\nGitOrigin-RevId: 09cd2944821fa43d97d8259194b9a0c4fa22de16\n" }, { "commit": "577075d233e61808edf5bb2a3ac149dc62ba2529", "tree": "8e3ee852cee62a1db0f20f48df3e16b63fee7815", "parents": [ "a4ae52b3837d3f2e338a63d790122aea05fae12c" ], "author": { "name": "Julian Brown", "email": "julian.brown@amd.com", "time": "Mon Apr 27 10:34:35 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Apr 27 02:35:30 2026 -0700" }, "message": "[OpenMP] Make loop index unsigned in __kmpc_omp_task_with_deps/__kmp_omp_task\n\nNFC.\n\nCo-authored-by: Adrian Munera \u003cadrian.munera@bsc.es\u003e\n\nReviewers: ro-i\n\nPull Request: https://github.com/llvm/llvm-project/pull/194044\n\nGitOrigin-RevId: 295a7a9f75b839394b49d4ec59b2df33b30cd4fd\n" }, { "commit": "a4ae52b3837d3f2e338a63d790122aea05fae12c", "tree": "f2c4db5ed29e8b62298104531529e8a5bc75c911", "parents": [ "091b48bfe86c0450ac658713d354fc181c9fc53c" ], "author": { "name": "nvptm", "email": "pmathew@nvidia.com", "time": "Thu Apr 23 15:13:34 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 23 15:16:04 2026 -0700" }, "message": "[flang] Add comparison operators for c_devptr (#192687)\n\nGitOrigin-RevId: 8b96c2104e74f90d6431022001cac2dccdbdeed0\n" }, { "commit": "091b48bfe86c0450ac658713d354fc181c9fc53c", "tree": "987bb5506f947dba246ac774a90c6739be04618f", "parents": [ "e32a0cf884cbe182a7392e1b59c5d2d086ddb49d" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Thu Apr 23 13:13:12 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 23 13:15:20 2026 -0700" }, "message": "[NFC][offload][OpenMP] Fix kernel replay documentation (#193832)\n\nGitOrigin-RevId: 3baafed3e779afc9c60149cd2528d70b81f3ff29\n" }, { "commit": "e32a0cf884cbe182a7392e1b59c5d2d086ddb49d", "tree": "8c9bffb82b5f4562c9fdf6798fd98628163570b6", "parents": [ "92377c5c346d3d0a5f161dbd5d43d13898e4588b" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Thu Apr 23 09:44:06 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 23 09:45:32 2026 -0700" }, "message": "[offload][OpenMP] Add basic documentation for kernel record replay (#193699)\n\nGitOrigin-RevId: 254fcbeface8716a42dfd38ded3a65dc0d72a55b\n" }, { "commit": "92377c5c346d3d0a5f161dbd5d43d13898e4588b", "tree": "4c036d3ea1a7ee21f8e877e68d2338115cc1eae5", "parents": [ "17568f01d137f57cd9646fb46a55c629be200669" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Thu Apr 23 09:42:34 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 23 09:45:25 2026 -0700" }, "message": "[offload] Fix envar description in docs (#193642)\n\nGitOrigin-RevId: df77a292371c3984d00816466b7b73ad3ebd21fb\n" }, { "commit": "17568f01d137f57cd9646fb46a55c629be200669", "tree": "04bb835accbabb807718a33948c5e3c8688d67c0", "parents": [ "e5e2a00d50c181c17797252737ea58c216456f63" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed Apr 22 07:46:12 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Apr 22 05:47:33 2026 -0700" }, "message": "[OpenMP] Fix OpenMP device subdir installation w/ multilibs (#193378)\n\nSummary:\nThese did not use the correct subdir so if you did a mulitilib build of\nthe OpenMP runtime it would simply overwrite it.\nGitOrigin-RevId: 58127f3ebffde72d942d6c772779519975cf006b\n" }, { "commit": "e5e2a00d50c181c17797252737ea58c216456f63", "tree": "692276b050ad58b4af241e911a4c3a6e303a94e1", "parents": [ "ac746973391bced14fa7ec18551bdb4afc95cf48" ], "author": { "name": "Nick Sarnie", "email": "nick.sarnie@intel.com", "time": "Fri Apr 17 03:51:52 2026 +0900" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 16 11:55:36 2026 -0700" }, "message": "[OpenMP][Device] Fix __llvm_omp_indirect_call_lookup function pointer types (#192502)\n\n`__llvm_omp_indirect_call_lookup` takes in and returns a function\npointer, so make sure the types are correct, which includes the correct\naddress space.\n\nThe FE was recently changed to generate the correct code\n[here](https://github.com/llvm/llvm-project/pull/192470).\n\nWith this change, three function pointer tests start passing.\n\nSigned-off-by: Nick Sarnie \u003cnick.sarnie@intel.com\u003e\nGitOrigin-RevId: 5aea02a163d94092b54833001e06a26c7e57fb01\n" }, { "commit": "ac746973391bced14fa7ec18551bdb4afc95cf48", "tree": "d1d000c0b0258d4f1bddcd14279cfb2ecb69e4d2", "parents": [ "aa5e99eff8f373defb27b46b639a48f6cb2d9aa9" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Wed Apr 15 17:01:51 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Apr 15 17:05:48 2026 -0700" }, "message": "[offload][OpenMP] Require CUDA 11.8 (#191100)\n\nGitOrigin-RevId: 18bed37e5ddadb1bf7a3439b746deb43a23699d5\n" }, { "commit": "aa5e99eff8f373defb27b46b639a48f6cb2d9aa9", "tree": "f357b61d487f1f4306662bcdc835aa7659e9d80f", "parents": [ "0986abcb14f126d4ab0d3452e6adb5cc7263f048" ], "author": { "name": "Alexandre Ganea", "email": "aganea@havenstudios.com", "time": "Wed Apr 15 07:15:14 2026 -0400" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Apr 15 04:20:56 2026 -0700" }, "message": "[openmp] Fix Darwin after 988e00e (#192135)\n\nAs per\nhttps://github.com/llvm/llvm-project/pull/191556#issuecomment-4246459320\n\nGitOrigin-RevId: 40690e55a8bac17d71604b651562985051a05cd9\n" }, { "commit": "0986abcb14f126d4ab0d3452e6adb5cc7263f048", "tree": "1b3a66201c060db32866b01cde4b2acf67105559", "parents": [ "901bca81cbbfd88868c1b290cac5a79625a24fc3" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed Apr 15 10:42:28 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 14 19:45:50 2026 -0700" }, "message": "[OpenMP] Create check-openmp target for device targets (#192175)\n\noffload/cmake/caches/AMDGPUBot.cmake enables\nRUNTIMES_amdgcn-amd-amdhsa_LLVM_ENABLE_RUNTIMES\u003d\"openmp\". In that\nsub-build, check-openmp target doesn\u0027t exist and there is build error\n`unknown target \u0027check-openmp\u0027` after 18f63d1375d0, which makes\ntop-level check-openmp depend on check-openmp-amdgcn-amd-amdhsa.\n\nIn openmp, the device targets only call add_subdirectory(device), which\ndoesn\u0027t calls construct_check_openmp_target() and check-openmp target\ndoesn\u0027t exist. `ninja check-openmp-amdgcn-amd-amdhsa` also fails with\nthe same error before 18f63d1375d0.\n\nFix by adding construct_check_openmp_target() for device targets as well.\n\nAssisted-by: Claude Sonnet 4.6\nGitOrigin-RevId: 9b8611b1c5b3d454b2c5e052c00f7285d733d496\n" }, { "commit": "901bca81cbbfd88868c1b290cac5a79625a24fc3", "tree": "8486522d6ce8de754ba9a432338dba96d93cd63c", "parents": [ "156356d00a3d8a429bc6f1b729b8345fb9df7d9a" ], "author": { "name": "Jan André Reuter", "email": "j.reuter@fz-juelich.de", "time": "Tue Apr 14 16:54:57 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 14 08:00:25 2026 -0700" }, "message": "[OpenMP][OMPT] Revert `control_tool.c` changes from #191429 (#192069)\n\nOn s390x, the changes to `control_tool.c` cause a different return\naddress to be returned from the call to `print_current_address(0)`. Due\nto the strictness of the current address returned by this call, this\nlead to a test failure.\n\nSince the return values of `omp_control_tool` are checked in separate\ntests already, revert the changes to ensure that the return address\nstays at the expected value.\n\nSigned-off-by: Jan André Reuter \u003cj.reuter@fz-juelich.de\u003e\nGitOrigin-RevId: 1737a913ed015796c9ba843de6c410c727070022\n" }, { "commit": "156356d00a3d8a429bc6f1b729b8345fb9df7d9a", "tree": "99d00a4f6e7fa95d0b794e50566965688f3d5857", "parents": [ "11cf6d0c095b5ead0fd389ffc8f3fab5a65068d4" ], "author": { "name": "Ye Luo", "email": "yeluo@anl.gov", "time": "Mon Apr 13 21:39:32 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Apr 13 19:45:28 2026 -0700" }, "message": "[Offload] Revert part of #187138. Workaround #191910 (#191925)\n\nCloses #191910\n\n---------\n\nCo-authored-by: Joseph Huber \u003chuberjn@outlook.com\u003e\nGitOrigin-RevId: aab5c1075a94b8600f2b31dd62aa4f2e0cd238cb\n" }, { "commit": "11cf6d0c095b5ead0fd389ffc8f3fab5a65068d4", "tree": "31f7b29898e5279d8953308a5cfdec2b23705f2c", "parents": [ "3773baf396d2ea237e4acf4c25b11b5f64e7aac9" ], "author": { "name": "Alexandre Ganea", "email": "aganea@havenstudios.com", "time": "Mon Apr 13 07:11:35 2026 -0400" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Apr 13 04:15:27 2026 -0700" }, "message": "[openmp] Silence warnings when building on Windows (#191556)\n\nFixes unused-but-set globals on non-Unix paths in kmp_alloc.cpp\n\nGitOrigin-RevId: 988e00ebcd54db5c4a15d15b87c16313d4190b59\n" }, { "commit": "3773baf396d2ea237e4acf4c25b11b5f64e7aac9", "tree": "807c37da5895e1a3e07f9b43be5c25b307a3cd52", "parents": [ "503d75aff95e2f88e89e9c67a67d2f257b2ca747" ], "author": { "name": "Amit Tiwari", "email": "amtiwari@amd.com", "time": "Mon Apr 13 12:39:40 2026 +0530" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Apr 13 00:16:18 2026 -0700" }, "message": "[Clang][OpenMP] Implement Loop splitting `#pragma omp split` directive (#190397)\n\nImplement Loop-splitting #pragma omp split construct with counts clause.\nPosting this PR after the revert of PR\n([#183261](https://github.com/llvm/llvm-project/pull/183261))\n\nChanges:\n\n1. Added `openmp/runtime/test/transform/split/lit.local.cfg`\n2. Enforced ICE for `counts` clause items in `SemaOpenMP.cpp` (minor\nchange)\n3. Updated tests `split_messages.cpp`, `split_omp_fill.cpp`,\n`split_diag_errors.c`.\n4. Removed `nonconstant_count.cpp`\n\nGitOrigin-RevId: 1d18740d307bb30cca9e9c8b2dc15e161d8e10f0\n" }, { "commit": "503d75aff95e2f88e89e9c67a67d2f257b2ca747", "tree": "65c7f4192d6507eb3403452b31ca004e762cd22f", "parents": [ "eaa73c92e09c6586917c02e35d6660cf6b259e6a" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Sat Apr 11 15:17:42 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Apr 11 06:20:29 2026 -0700" }, "message": "[OpenMP] Fix nondependent inscan variables in templated functions (#191627)\n\nFixes https://github.com/llvm/llvm-project/issues/191549.\n\nAssisted-by: claude-4.6-opus\nGitOrigin-RevId: 54d4bf2acbf4352e3c1751cd31211c925a041065\n" }, { "commit": "eaa73c92e09c6586917c02e35d6660cf6b259e6a", "tree": "1b287f80b94adc0df992f5878a364293416e9b39", "parents": [ "79db276505d0509aff7704720099448b870d642d" ], "author": { "name": "Jan André Reuter", "email": "j.reuter@fz-juelich.de", "time": "Fri Apr 10 17:07:07 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Apr 10 08:11:57 2026 -0700" }, "message": "[OpenMP][OMPT] Fix `omp_control_tool` before any directive (#191429)\n\nWhen a user calls `omp_control_tool`, a tool is attached and it\nregistered the `ompt_control_tool` callback, the tool should receive a\ncallback with the users arguments.\n\nHowever, in #112924, it was discovered that this only happens after at\nleast one host side directive or runtime call calling into\n`__kmp_do_middle_initialize` has been executed.\n\nThe check for `__kmp_init_middle` in `FTN_CONTROL_TOOL` did not try to\ndo the middle initialization and instead always returned `-2` (no tool).\nA tool therefore received no callback. The user program did not get the\ninfo that there is a tool attached. To fix this, change the explicit\nreturn to a call of `__kmp_middle_initialize()`, as done in several\nother places of `libomp`.\n\nFurther handling is then done in `__kmp_control_tool`, where the values\n`-2` (no tool), `-1` (no callback), or the tools return value are\nreturned.\n\nAlso expand the tests to introduce checks where no callaback is\nregistered, or `omp_control_tool` is called before any OpenMP directive.\n\nFixes #112924\n\nCC @jprotze, @hansangbae\n\nSigned-off-by: Jan André Reuter \u003cj.reuter@fz-juelich.de\u003e\nGitOrigin-RevId: c6bcd194259e276aacadae8ac0d25680afc9489d\n" }, { "commit": "79db276505d0509aff7704720099448b870d642d", "tree": "786a56ce14ec6cb760bc0902738a246d170c1ea4", "parents": [ "9c7507e1b01b8431e1d37d19d58f99e88d186d2e" ], "author": { "name": "theRonShark", "email": "ron.lieberman@amd.com", "time": "Fri Apr 03 08:27:07 2026 -0400" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Apr 03 05:30:36 2026 -0700" }, "message": "Revert \"[Clang][OpenMP] Implement Loop splitting `#pragma omp split` directive \" (#190335)\n\nReverts llvm/llvm-project#183261\n\n15 new lit tests failing in openmp\n\nGitOrigin-RevId: 00aede8f19c1714e637ec65c9cf15e4f5559571a\n" }, { "commit": "9c7507e1b01b8431e1d37d19d58f99e88d186d2e", "tree": "cf847a1c2fee4357e09d6918de30428f31d13f4b", "parents": [ "e134bfb845c44a0dca14852c5888eb6ae3a7e559" ], "author": { "name": "Amit Tiwari", "email": "amtiwari@amd.com", "time": "Fri Apr 03 10:42:31 2026 +0530" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 02 22:15:21 2026 -0700" }, "message": "[Clang][OpenMP] Implement Loop splitting `#pragma omp split` directive (#183261)\n\nOpenMP 6.0 Loop-splitting directive `#pragma omp split` construct with `counts`\nclause\n\nGitOrigin-RevId: 1972cf64fd18227bd8c3cc94ad96164cae9a163a\n" }, { "commit": "e134bfb845c44a0dca14852c5888eb6ae3a7e559", "tree": "786a56ce14ec6cb760bc0902738a246d170c1ea4", "parents": [ "e9f45e5291646c6a91c5162b0e5bd3ba15e0190a" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Thu Apr 02 12:32:14 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 02 03:35:07 2026 -0700" }, "message": "[Runtimes] Introduce variables containing resource dir paths (#177953)\n\nIntroduce common infrastructure for runtimes that determines compiler\nresource path locations. These variables introduced are:\n\n * RUNTIMES_OUTPUT_RESOURCE_DIR\n * RUNTIMES_INSTALL_RESOURCE_PATH\n\nThat contain the location for the compiler resource path (typically\n`lib/clang/\u003cversion\u003e`) in the build tree and the install tree (the\nlatter relative to CMAKE_INSTALL_PREFIX).\n\nAdditionally, define\n\n * RUNTIMES_OUTPUT_RESOURCE_LIB_DIR\n * RUNTIMES_INSTALL_RESOURCE_LIB_PATH\n\nas for the location of clang/flang version-locked libraries (typically\n`lib${LLVM_LIBDIR_SUFFIX}/\u003ctarger-triple\u003e`, but also depends on `APPLE`\nand `LLVM_ENABLE_PER_TARGET_RUNTIME_DIR`). This code is moved from\nflang-rt and initially becomes its only user.\n\nRefactored out of #171610 as requested\n[here](https://github.com/llvm/llvm-project/pull/171610#discussion_r2687382481).\n\nExtracted `get_runtimes_target_libdir_common` from compiler-rt as\nrequested\n[here](https://github.com/llvm/llvm-project/pull/171610#discussion_r2689565634).\n\nAdded TODO comments to all runtimes as requested\n[here](https://github.com/llvm/llvm-project/pull/171610#issuecomment-3789598635).\n\nGitOrigin-RevId: afb80bddf147c134c5e3cf3f357f8d00bdfd526c\n" }, { "commit": "e9f45e5291646c6a91c5162b0e5bd3ba15e0190a", "tree": "4238547391102e841021790f3b7a44e99c880e86", "parents": [ "d3d82ac49743d4379bcf9cf1abc0048af283b48c" ], "author": { "name": "David Truby", "email": "david.truby@arm.com", "time": "Tue Mar 31 15:39:02 2026 +0000" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 31 08:40:39 2026 -0700" }, "message": "[openmp] Add support for Arm64X to libomp (#176157)\n\nThis patch allows building libomp.dll and libomp.lib as Arm64X binaries\ncontaining both arm64 and arm64ec code and useable from applications\ncompiled for both architectures.\n\nGitOrigin-RevId: 720b8ea893da2168121622bbfb536a78b35ffa32\n" }, { "commit": "d3d82ac49743d4379bcf9cf1abc0048af283b48c", "tree": "efbbc08d3b67ab4f58e9ab1620dcb5bba462549b", "parents": [ "6745aa3941f080aefdbba408ae094df39be75f16" ], "author": { "name": "David Truby", "email": "david.truby@arm.com", "time": "Tue Mar 31 13:07:19 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 31 05:11:03 2026 -0700" }, "message": "[openmp] Add support for arm64ec to libomp (#176151)\n\nThis patch adds arm64ec support to libomp.\nNote that this support isn\u0027t entirely usable on Windows hosts as libomp\nrequires LLVM_PER_TARGET_RUNTIME_DIR\u003dOn for to work correctly when\nmultiple runtimes are built, which is unsupported on Windows. A\nfollowing patch will add arm64x support to the build to rectify this.\n\nGitOrigin-RevId: 161c53f66ea8044e6618d7c7829ba90196c72a79\n" }, { "commit": "6745aa3941f080aefdbba408ae094df39be75f16", "tree": "c648ab27cde5f03591703cd3ce0a1bdd142b9e75", "parents": [ "b6246e910eaab5ba9cc84d8d707ad83b0e0c40a8" ], "author": { "name": "fineg74", "email": "61437305+fineg74@users.noreply.github.com", "time": "Fri Mar 27 12:14:09 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 12:15:20 2026 -0700" }, "message": "[OFFLOAD] Add spirv implementation for named barrier (#180393)\n\nThis change adds implementation for named barriers for SPIRV backend.\nSince there is no built in API/intrinsics for named barrier in SPIRV,\nthe implementation loosely follows implementation for AMD\n\nGitOrigin-RevId: 1611a23a5b34c0faf464d5ce605fe2d4a86baaa6\n" }, { "commit": "b6246e910eaab5ba9cc84d8d707ad83b0e0c40a8", "tree": "f63bf89843fe1c07c031f62ee77435f1e2eb50c3", "parents": [ "7c7044536180072817c4dcb1ca5a2f073bff9a0e" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Fri Mar 27 10:59:06 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 03:00:51 2026 -0700" }, "message": "[OpenMP][test] Remove %flags-use-compiler-omp-h (#188789)\n\nWith the standalone and project builds removed,\n`OPENMP_TEST_COMPILER_HAS_OMP_H`/`config.test_compiler_has_omp_h` is set\nto constant 1, which causes the `config.omp_header_directory` search\npath NOT to be added to `%flags-use-compiler-omp-h`, causing the system\n`omp.h` used, or the only test actually using it\n(`omp50_taskdep_depobj.c`) failing if that one is not available.\n\nThe intention of `OPENMP_TEST_COMPILER_HAS_OMP_H` was to use gcc\u0027s\n`omp.h` which declares `omp_depend_t` differently than our `omp.h`\n(https://reviews.llvm.org/D108790). Using `OPENMP_TEST_C_COMPILER\u003dgcc`\nwas used to test libomp\u0027s GOMP compatibility layer, but testing it is\ncurrently unmaintained and has no buildbot (60 failing tests out of 389\nwith gcc-13, not including OMPD and OMPT). If updating testing for GOMP,\nthen gcc\u0027s own `omp.h` must be used for all tests: using the GOMP ABI\nrequires using GOMP\u0027s `omp.h`.\n\nCloses: #187879\nGitOrigin-RevId: 8ab7b37e749f4ab7c4536fec82cb813586d9f334\n" }, { "commit": "7c7044536180072817c4dcb1ca5a2f073bff9a0e", "tree": "2d21201a51d670ba66dac0d90ee7bbd2bfffe111", "parents": [ "7af3d7bb6da3f6edbd5432993158bb24eca499e2" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu Mar 26 07:37:22 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 26 05:41:06 2026 -0700" }, "message": "[Offload] Enable multilib building for OpenMP/Offload (#188485)\n\nSummary:\nRight now the CMake does not follow the pattern other runtime projects\nuse. All this does is use the standard subdir to place libraries in a\nunique location. This allows, for example, users to configure a debug\nversion of openmp / offload within the same CMake invocation.\n\n---------\n\nCo-authored-by: Michael Kruse \u003cgithub@meinersbur.de\u003e\nGitOrigin-RevId: 82530154ef4bb0c454170c6d420bfbcaef4e1dc4\n" }, { "commit": "7af3d7bb6da3f6edbd5432993158bb24eca499e2", "tree": "7e7b62ffc82a2ef040fefa9cfc8afa533f761e3b", "parents": [ "b8c79b756bc0b26a81a591e4679cda80a4a211ba" ], "author": { "name": "Alex Duran", "email": "alejandro.duran@intel.com", "time": "Wed Mar 25 16:22:22 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 08:25:34 2026 -0700" }, "message": "[Offload][OpenMP][libdevice] Make check to enter state machine architecture dependent (#188144)\n\nThe genericStateMachine call uses synchronize::thread wich is expected\nto be implemented using a workgroup level barrier.\nCurrently as in some other architectures where if threads in the same\nwarp as the main thread reach the barrier may cause a race condition\nthere\u0027s a condition that makes some threads not enter the state machine.\nBut in Intel GPUs all threads must reach the barrier for it to be\ncompleted, otherwise the threads in the state machine never make\nprogress.\n\nThis PR moves the condition into an architecture-dependent config so it\ncan work correctly for both kinds of hardware.\n\nGitOrigin-RevId: 3e72f02e20b99ba1fa0dce2982d89aaa0ef6ab26\n" }, { "commit": "b8c79b756bc0b26a81a591e4679cda80a4a211ba", "tree": "9ed4e759802fb5d3a0d29f7efde577c99227f6ff", "parents": [ "1ac7402562c06be6fa9ea65ea2719bdc46543846" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu Mar 19 10:51:48 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 19 08:56:20 2026 -0700" }, "message": "[compiler-rt] Define GPU specific handling of profiling functions (#185763)\n\nSummary:\nThe changes in https://www.github.com/llvm/llvm-project/pull/185552\nallowed us to\nstart building the standard `libclang_rt.profile.a` for GPU targets.\nThis PR expands this by adding an optimized GPU routine for counter\nincrement and removing the special-case handling of these functions in\nthe OpenMP runtime.\n\nVast majority of these functions are boilerplate, but we should be able\nto do more interesting things with this in the future, like value or\nmemory profiling.\n\nGitOrigin-RevId: d18a784d410617fbb3f9d2bf6d30fcd84b3533db\n" }, { "commit": "1ac7402562c06be6fa9ea65ea2719bdc46543846", "tree": "ce684845f0310da648ab027fd26ca3b6d3189b66", "parents": [ "fbf0084b0c5ae2367573f6a02c1511aebb424371" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed Mar 18 08:42:07 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 18 06:45:36 2026 -0700" }, "message": "[OpenMP] Use the standard fences now that they are supported (#187138)\n\nSummary:\nWe can now use these on NVPTX properly.\nGitOrigin-RevId: 073d019c450da7948a23ae240bfc6388971a8f25\n" }, { "commit": "fbf0084b0c5ae2367573f6a02c1511aebb424371", "tree": "7054faa2311673211f3d822b391937e5a600d11b", "parents": [ "158047468687e92f8a6ef2b9c806db5655978a1b" ], "author": { "name": "Jan André Reuter", "email": "j.reuter@fz-juelich.de", "time": "Tue Mar 17 10:00:21 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 17 02:03:22 2026 -0700" }, "message": "[OpenMP][OMPT] Add missing `error` entry to device tracing record union (#185683)\n\nWhile `omp-tools.h` already includes the `ompt_record_error_t` struct,\nthe corresponding union entry was missing from `ompt_record_ompt_t`.\nThis commit adds the missing entry.\n\nNote that this does not enable any functionality for device tracing\nrecords.\nThis only aligns the struct with OpenMP v5.1 and newer. OpenMP v5.0 did\nnot contain the `error` directive.\n\nCC: @jprotze\n\nSigned-off-by: Jan André Reuter \u003cj.reuter@fz-juelich.de\u003e\nGitOrigin-RevId: 1588f083694b9547e974f551b4e3110c35b1dc4d\n" }, { "commit": "158047468687e92f8a6ef2b9c806db5655978a1b", "tree": "f69267a185843ad4d0c73b1856ddfebfa53a7a45", "parents": [ "452e5b6dd540e844a77318a13419882632e1e122" ], "author": { "name": "Chi-Chun, Chen", "email": "chichun.chen@hpe.com", "time": "Mon Mar 16 10:16:38 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Mar 16 08:20:29 2026 -0700" }, "message": "[mlir][llvmir][OpenMP] Translate affinity clause in task construct to llvmir (#182223)\n\nTranslate affinity entries to LLVMIR by passing affinity information to\ncreateTask (__kmpc_omp_reg_task_with_affinity is created inside\nPostOutlineCB).\n\n3/3 in stack for implementing affinity clause with iterator modifier\n1/3 #182218\n2/3 #182222\n3/3 #182223\n\nGitOrigin-RevId: 2ad51ffbfae77ecb67e64ed8e7e833e54285f4c2\n" }, { "commit": "452e5b6dd540e844a77318a13419882632e1e122", "tree": "ae67eea9f009d8d74541a25c0485a5466dff450f", "parents": [ "83f638d3f049a439bb7f301417df4a595788f437" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Thu Mar 12 21:21:29 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 12 21:25:22 2026 -0700" }, "message": "[offload] Remove LIBOMPTARGET_SHARED_MEMORY_SIZE envar (#186231)\n\nThis commit removes the `LIBOMPTARGET_SHARED_MEMORY_SIZE` envar and\noutputs a runtime warning if it is defined. Access to dynamic shared memory\nshould be obtained through the `dyn_groupprivate` clause (OpenMP 6.1) or\nthe launch arguments in liboffload kernel launch.\n\nGitOrigin-RevId: ac71b185c2fd6b23e107a29f630d3a927ab6b3b0\n" }, { "commit": "83f638d3f049a439bb7f301417df4a595788f437", "tree": "08a42bc9f3edd0d0f87ef232a9d6ab635cc40ada", "parents": [ "7d044a1df58ead0b38bbd788c273dca015d6d4b7" ], "author": { "name": "Michael Halkenhäuser", "email": "MichaelGerald.Halkenhauser@amd.com", "time": "Thu Mar 12 18:16:36 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 12 10:20:28 2026 -0700" }, "message": "[OpenMP][OMPT] Remove Threads dependency from omptest link interface (#186183)\n\nLink against Threads using PRIVATE scope, instead of PUBLIC.\nReason: it imposes a transitive dependency on library users.\n\nIf Threads could not be found this could cause a link error.\nThe issue would manifest, if omptest is used via find_package.\n\nAddresses issues with previous PR\nhttps://github.com/llvm/llvm-project/pull/185930\n\nGitOrigin-RevId: 4f2dea8ac71485cfb2092a0671c08946b3ad31eb\n" }, { "commit": "7d044a1df58ead0b38bbd788c273dca015d6d4b7", "tree": "f7576e5696ea8164ea84e50e0e78b75f860edc42", "parents": [ "a06bc8eae4c35b4c6e65c0ba302c7eaf002aa606" ], "author": { "name": "Kelvin Li", "email": "kli@ca.ibm.com", "time": "Thu Mar 12 11:54:11 2026 -0400" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 12 08:55:06 2026 -0700" }, "message": "[OpenMP] Include the cmake policy file (#186031)\n\nThis patch is to enforce building shared library archive by default on\nAIX as described in CMP0182.\n\nGitOrigin-RevId: 9ddf38b008c3d87dd5fec47565fcde56180c92f6\n" }, { "commit": "a06bc8eae4c35b4c6e65c0ba302c7eaf002aa606", "tree": "af6cd8afbfbed2c0dc8bf3b7baf984f92436be7a", "parents": [ "97d9371b3aabe8073d8f027742502ff775d728ad" ], "author": { "name": "Jan Patrick Lehr", "email": "JanPatrick.Lehr@amd.com", "time": "Thu Mar 12 13:51:01 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 12 05:53:22 2026 -0700" }, "message": "Revert \"[OpenMP][OMPT] Remove Threads dependency from omptest\" (#186111)\n\nReverts llvm/llvm-project#185930\n\nBreaks various buildbots\n\nGitOrigin-RevId: 1cd094fb704de54fbc3d33456f9ec9bf5bcfb9df\n" }, { "commit": "97d9371b3aabe8073d8f027742502ff775d728ad", "tree": "de964112d8447626e2046135faf4e5dede6748c5", "parents": [ "0bd4d7a74ba60467ed0558f10c03db0e1915595a" ], "author": { "name": "Michael Halkenhäuser", "email": "MichaelGerald.Halkenhauser@amd.com", "time": "Thu Mar 12 13:08:49 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 12 05:10:50 2026 -0700" }, "message": "[OpenMP][OMPT] Remove Threads dependency from omptest (#185930)\n\nRemoved link against `Threads`.\nReason: it is potentially problematic and optional.\n\nThe issue would manifest, if `omptest` is used via `find_package`.\nBut `Threads` might not be found and cause a link error.\n\nGitOrigin-RevId: 5f97e191b34f4fd6185a9caa7468aa0432c5d503\n" }, { "commit": "0bd4d7a74ba60467ed0558f10c03db0e1915595a", "tree": "af6cd8afbfbed2c0dc8bf3b7baf984f92436be7a", "parents": [ "7dfb97ad126eae025412f974ff1587fa0ce92a1d" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Thu Mar 12 01:13:06 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 12 01:16:03 2026 -0700" }, "message": "[OpenMP][Offload] Add offload runtime support for dyn_groupprivate clause (#152831)\n\nPart 3 adding offload runtime support. See\nhttps://github.com/llvm/llvm-project/pull/152651.\n\n---------\n\nCo-authored-by: Krzysztof Parzyszek \u003cKrzysztof.Parzyszek@amd.com\u003e\nGitOrigin-RevId: 1f583c6dee360b0f5837a1026f2c594643cf885c\n" }, { "commit": "7dfb97ad126eae025412f974ff1587fa0ce92a1d", "tree": "aca8cce572415c06e6056950ed34cbf8c537dfc3", "parents": [ "d915830a7705416436c9f61d89ff9e1deadbcde0" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Mon Mar 02 18:03:36 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Mar 02 09:05:18 2026 -0800" }, "message": "[libomp] Fix hwloc include for non-standard paths (#184087)\n\nFixes https://github.com/llvm/llvm-project/issues/183884\n\n---------\n\nCo-authored-by: Michael Kruse \u003cllvm-project@meinersbur.de\u003e\nGitOrigin-RevId: fd578f7c5c98d7a5edfbd1ad5f72db6541ddc679\n" }, { "commit": "d915830a7705416436c9f61d89ff9e1deadbcde0", "tree": "a16a93695abba059b8d149c8f67f04ecc89ad4fc", "parents": [ "3e9eaf6f2963a4b9305c50d1a4dd7bc68ebd1135" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Mon Mar 02 13:21:36 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Mar 02 04:26:19 2026 -0800" }, "message": "[openmp][cmake][NFCI] Avoid non-eval uses of ${var} (#182267)\n\nWhen using\n\n set(var \"Example\")\n if (${var})\n\nCMake will first resolve the if-argument to\n\n if (Example)\n\nAnd what then happens depends on the existance of a variable with the\nname \"Example\":\n\n1. If instead of \"Example\", a truth constant is used (\"TRUE\", \"FALSE\",\n \"ON\", \"OFF\", \"\", \"-NOTFOUND\" ...), it is prioritized\n (https://cmake.org/cmake/help/latest/command/if.html#constant)\n\n2. If a variable with the name \"Example\" exists: Use the truthiness of\n its content\n (https://cmake.org/cmake/help/latest/command/if.html#variable)\n\n3. Otherwise, it is false-y\n\nThat is, the the result of the conditional does not only depend on the\ncontent of `var`, but also some other variable. This is usually\nunintended and leads to problems such as addressed with #154537. The\nonly case where this is intended is when passing an expression to be\nevaluated such as with `pythonize_bool`, `append_if` and\n`libomp_append`. In all other cases, using `${var}` without quotes is a\npattern to be avoided.\n\nRemark:\nIf `${var}` is not one of the values considered a [truthiness\nconstant](https://cmake.org/cmake/help/latest/command/if.html#constant),\nthe result of `if (var)` and `if (\"${var}\")` is different:\n\n* `if (var)` is true-ish\n (https://cmake.org/cmake/help/latest/command/if.html#variable)\n\n* `if (\"Example\")` and therefore `if (\"${var}\")` are false-y\n (https://cmake.org/cmake/help/latest/command/if.html#string)\n\nIn this PR I am preferring `if (var)` over `if (\"${var}\")` because has\nless clutter, resembles Python\u0027s behaviour, and problably what most\nusers are expecting, even though `if (${var})` in most cases would\nevaluate to false-y because a variable does not exist. This ambiguity\ndoes not exist for STREQUAL and MATCHES.\n\nGitOrigin-RevId: 87cbea6cdc9a55c1fda2d83820d0d81f314868a7\n" }, { "commit": "3e9eaf6f2963a4b9305c50d1a4dd7bc68ebd1135", "tree": "88a2ca312cb600526ac56881208c5c0923cdbdc8", "parents": [ "d26b5471bf0ee5e8af31e6de1f2e37aa4152e246" ], "author": { "name": "Joachim", "email": "jenke@itc.rwth-aachen.de", "time": "Thu Feb 26 17:26:47 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Feb 26 08:30:06 2026 -0800" }, "message": "[openmp][tests] Fix bug63197.c (#183508)\n\n#183269 tried to fix the test, but the test can still randomly fail. The\nOpenMP spec does not prevent the runtime to chose a smaller team size\nthan returned from omp_max_threads() for the second parallel region.\n\nUsing a larger value than `omp_max_threads()` in a `num_threads` clause\nis valid OpenMP code. With a correct OpenMP implementation, the\nteam-size of the second parallel region must still be smaller or equal\nthe value returned from `omp_max_threads()`.\n\nGitOrigin-RevId: d44e41794540d7fa85f857228a7616ec8592a7c4\n" }, { "commit": "d26b5471bf0ee5e8af31e6de1f2e37aa4152e246", "tree": "a12dfeb9cb13067e65df4019c6bbfa9f205870e4", "parents": [ "f0dc530923c422ab73d686d5587e3ec8994b955e" ], "author": { "name": "Nikita Popov", "email": "npopov@redhat.com", "time": "Wed Feb 25 15:58:23 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Feb 25 07:00:24 2026 -0800" }, "message": "[openmp] Fix bug63197.c test with 3 cores (#183269)\n\nThis test assumes that the number of available threads is not 3,\notherwise `#pragma omp parallel` and `#pragma omp parallel\nnum_thread(3)` are naturally going to do the same thing.\n\nInstead use `omp_get_max_threads() - 1` as the number of threads in the\ninitial `omp parallel num_thread(N)` and then check that the number of\nthreads does not match the value in the later `omp parallel`.\n\nGitOrigin-RevId: 368b884869651b69b14b836b25206aa62b501496\n" }, { "commit": "f0dc530923c422ab73d686d5587e3ec8994b955e", "tree": "61f6c0290953268c15c50db28052adc3b6d4642e", "parents": [ "b1e4dfc8ef40ef7161a8c490ae963cbf76df8bcc" ], "author": { "name": "Martin Storsjö", "email": "martin@martin.st", "time": "Mon Feb 23 23:01:44 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Feb 23 13:07:57 2026 -0800" }, "message": "[openmp] Allow testing OpenMP without a full clang build tree (#182470)\n\nHaving a build tree with \"not\" and \"FileCheck\" is still required, but if\nClang/Flang isn\u0027t configured in that build, run the tests with the same\ncompiler CMake uses. This is how testing worked in the standalone build\nconfigurations that now have been removed.\n\nGitOrigin-RevId: 48a5119d8e7d7236a28c14d06ec215ef3366ef90\n" }, { "commit": "b1e4dfc8ef40ef7161a8c490ae963cbf76df8bcc", "tree": "26fe7d9eac4c306eaabc52c54f8034559d5b5420", "parents": [ "7c0ef25c407f080247923679ace1b6ef0bbde404" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Thu Feb 19 00:11:55 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Feb 18 15:15:57 2026 -0800" }, "message": "[OpenMP] Remove standalone build mode (Reapply #149878) (#182022)\n\nThis reapplies #149878\n\nRemove all the CMake code for openmp standalone builds. Standalone\nbuilds have been superseded by the runtimes default build (also\nsometimes called the standalone runtimes build). The runtimes default\nbuild can be thought of a standalone build with the standalone\nboilerplate contained in \u003cllvm-project\u003e/runtimes/CMakeLists.txt. There\nis no need for each runtime to contain the same boilerplate code again.\n\nBuilds still using the standalone build via\n```sh\ncmake -S \u003cllvm-project\u003e/openmp ...\n```\ncan switch over to the runtimes default build using\n```sh\ncmake -S \u003cllvm-project\u003e/runtimes -DLLVM_ENABLE_RUNTIMES\u003dopenmp ...\n```\nOptions that were valid for the standalone build are also valid for\ndefault runtimes build, unless handled only in `if\n(OPENMP_STANDALONE_BUILD)` regions.\n\nThe existence of both, standalone builds and runtime default builds,\neasily leads to confusion such as fixed in #149871. A non-standalone\nbuild does not mean that it is built as part of a bootstrapping-build of\nLLVM inside the `runtimes/runtimes-bins` directory, nor does it mean\nthat the compiler is necessarily Clang. Some of the remaining code may\nhave been written with that assumption. However, the purpose of this\npatch is to only remove the standalone build functionality.\n\nGitOrigin-RevId: 6295b8e763503644961f9087382f80965b98d466\n" }, { "commit": "7c0ef25c407f080247923679ace1b6ef0bbde404", "tree": "32b07fa3ceb8ab33257d38403b7833aec5ce3e1d", "parents": [ "0fa4b54e8ffdb54c468092773260a454dca4bf7f" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Wed Feb 18 13:52:11 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Feb 18 05:01:09 2026 -0800" }, "message": "Revert \"[OpenMP] Remove standalone build mode (#149878)\"\n\nDownstream builds are note sufficiently prepared.\n\nThis reverts commit 176928cc1371c23d499b474fa95e076b61f776c4.\n\nGitOrigin-RevId: e07b9d15043271fb69177cdc5df9c504a21764ec\n" }, { "commit": "0fa4b54e8ffdb54c468092773260a454dca4bf7f", "tree": "26fe7d9eac4c306eaabc52c54f8034559d5b5420", "parents": [ "c9b043fddad198fada5934a446467f3cb2a6b07b" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Wed Feb 18 12:00:04 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Feb 18 03:05:36 2026 -0800" }, "message": "[OpenMP] Remove standalone build mode (#149878)\n\nRemove all the CMake code for openmp standalone builds. Standalone\nbuilds have been superseded by the runtimes default build (also\nsometimes called the standalone runtimes build). The runtimes default\nbuild can be thought of a standalone build with the standalone\nboilerplate contained in \u003cllvm-project\u003e/runtimes/CMakeLists.txt. There\nis no need for each runtime to contain the same boilerplate code again.\n\nBuilds still using the standalone build via\n```sh\ncmake -S \u003cllvm-project\u003e/openmp ...\n```\ncan switch over to the runtimes default build using\n```sh\ncmake -S \u003cllvm-project\u003e/runtimes -DLLVM_ENABLE_RUNTIMES\u003dopenmp ...\n```\nOptions that were valid for the standalone build are also valid for\ndefault runtimes build, unless handled only in\n`if (OPENMP_STANDALONE_BUILD)` regions.\n\nThe existence of both, standalone builds and runtime default builds,\neasily leads to confusion such as fixed in #149871. A non-standalone\nbuild does not mean that it is built as part of a bootstrapping-build of\nLLVM inside the `runtimes/runtimes-bins` directory, nor does it mean\nthat the compiler is necessarily Clang. Some of the remaining code may\nhave been written with that assumption. However, the purpose of this\npatch is to only remove the standalone build functionality.\n\nGitOrigin-RevId: 176928cc1371c23d499b474fa95e076b61f776c4\n" }, { "commit": "c9b043fddad198fada5934a446467f3cb2a6b07b", "tree": "32b07fa3ceb8ab33257d38403b7833aec5ce3e1d", "parents": [ "1f45021c92fa75cf0e0ee829a03b53a021148b14" ], "author": { "name": "Ferran Toda", "email": "ferran.todacasaban@bsc.es", "time": "Tue Feb 17 15:34:27 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Feb 17 06:36:16 2026 -0800" }, "message": "[MLIR][OpenMP] Add omp.fuse operation (#168898)\n\nThis patch is a follow-up from #161213 and adds the omp.fuse loop\ntransformation for the OpenMP dialect. Used for lowering a `!$omp fuse`\nin Flang.\n\nAdded Lowering and end2end tests.\n\nGitOrigin-RevId: f560e4cfb1f9d3c71d7617f6e54f54eeba9f3239\n" }, { "commit": "1f45021c92fa75cf0e0ee829a03b53a021148b14", "tree": "aea47453cea4e84b899e5e42c1b5a367284eb9ae", "parents": [ "fad344f1beb6979b69d92adc5b58c9980e3a5ad2" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Mon Feb 16 15:14:30 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Feb 16 13:16:10 2026 -0800" }, "message": "[libc] Replace RPC \u0027close()\u0027 mechanism with RAII handler (#181690)\n\nSummary:\nClosing ports was previously done manually, This makes the protocol more\nerror prone as unclosed ports will leak and eventually the locks will\nrun out. I believe the original fear was that the RAII portion would\nnegatively impact code generation but I have not noticed anything\nsignificant.\nGitOrigin-RevId: d85576d3685d2db72e52bf555125539af3211dda\n" }, { "commit": "fad344f1beb6979b69d92adc5b58c9980e3a5ad2", "tree": "695004912138c72c63f4d97a4dd484930ca35bdb", "parents": [ "0dcdab516f2564580b6c3a9012fc093cc2d6a821" ], "author": { "name": "Nick Sarnie", "email": "nick.sarnie@intel.com", "time": "Sat Feb 14 02:06:36 2026 +0900" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Feb 13 09:10:38 2026 -0800" }, "message": "[OpenMP][lit] Disable flaky test tasking/issue-94260-2.c (#176452)\n\nThe test sporadically fails on Linux. See\nhttps://github.com/llvm/llvm-project/issues/176451 for more info.\n\nSigned-off-by: Nick Sarnie \u003cnick.sarnie@intel.com\u003e\nGitOrigin-RevId: b217e5c5a035834a1fa11c092479a50f91d8f90d\n" }, { "commit": "0dcdab516f2564580b6c3a9012fc093cc2d6a821", "tree": "c4cee74dc9429a3b9850e98a8946164c96d0c644", "parents": [ "51ec7deba4c364e7b4e8cbc9db389919455df0c6" ], "author": { "name": "Nikita Popov", "email": "npopov@redhat.com", "time": "Fri Feb 13 16:11:31 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Feb 13 07:15:48 2026 -0800" }, "message": "[openmp] Remove perl mentions (#181371)\n\nThese tests were converted from Perl to Python in #95307, so there is no\nperl dependency here anymore. (I omitted an explicit mention of Python\nhere, as that\u0027s a general dependency anyway.)\n\nGitOrigin-RevId: 05ea779fcf475b7edd7e27c11966366be1aa01f5\n" }, { "commit": "51ec7deba4c364e7b4e8cbc9db389919455df0c6", "tree": "fc7802948eba887b2a5bc4fd33594e3780076716", "parents": [ "b991d6eb2c24afd3b3c87cd0ce2fcefb11d821a8" ], "author": { "name": "fineg74", "email": "61437305+fineg74@users.noreply.github.com", "time": "Fri Feb 13 04:55:44 2026 -0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Feb 13 05:00:44 2026 -0800" }, "message": "[OFFLOAD] Add support for SPIRV to ompx (#179849)\n\nAfter adding support to build device RTL for SPIRV, this PR will make an\nompx to use the DeviceRTL API when compiled for SPIRV\n\nGitOrigin-RevId: e0293fc7a55aa880d1bb67d31775bdbe85c672d1\n" }, { "commit": "b991d6eb2c24afd3b3c87cd0ce2fcefb11d821a8", "tree": "26024cf7a6cdc554703d72043e2e34846e4c99a5", "parents": [ "1b0ea1d97cf5e8b66f46bc75ccd5d6ef6b8cac48" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue Feb 10 10:21:16 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Feb 10 08:25:45 2026 -0800" }, "message": "[OpenMP] Remove special handling for SPIR-V uinc atomic (#180747)\n\nSummary:\nNo longer necessary after\nhttps://github.com/llvm/llvm-project/pull/179114\nGitOrigin-RevId: a969d572c27d49f9d8d4852fba798cfd41c76987\n" }, { "commit": "1b0ea1d97cf5e8b66f46bc75ccd5d6ef6b8cac48", "tree": "14cf2a28c008d204c47c45fbd7d050ad2ce914cb", "parents": [ "1fa3e4bb61a73d61c1a70c4a9d7130fe9d59183a" ], "author": { "name": "YunQiang Su", "email": "syq@debian.org", "time": "Fri Feb 06 08:17:56 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Feb 05 16:18:42 2026 -0800" }, "message": "openmp: add atomic to SANITIZER_COMMON_LINK_LIBS for MIPS (#179933)\n\natomic is needed explicitly for MIPS.\n\nGitOrigin-RevId: 74e9bc6cb93964edfcb3761623a43d1e55aceb2a\n" }, { "commit": "1fa3e4bb61a73d61c1a70c4a9d7130fe9d59183a", "tree": "9c9f61dad5873ba727471f719fe6b623f4f57158", "parents": [ "48d38978328b912c1dff5b83d87b49b16cb95b29" ], "author": { "name": "Nick Sarnie", "email": "nick.sarnie@intel.com", "time": "Fri Feb 06 07:11:32 2026 +0900" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Feb 05 14:15:31 2026 -0800" }, "message": "[OpenMP] Stub vprintf on non-NVPTX if no libc (#180025)\n\nThe AMDGPU check was added in\nhttps://github.com/llvm/llvm-project/pull/123670 where the reasoning\nseems to be that the NVIDIA SDK will provide `vprintf` for the NVPTX\ncase and AMDGPU was the only other supported target at the time.\n\nSPIR-V also needs this stubbed out, so just check that it\u0027s not NVPTX.\n\nSigned-off-by: Nick Sarnie \u003cnick.sarnie@intel.com\u003e\nGitOrigin-RevId: 02d8c9c63a9925cea2ab7560742dbf2619da89cc\n" }, { "commit": "48d38978328b912c1dff5b83d87b49b16cb95b29", "tree": "d0e9df4045879721e8469e18cbdede99147a9903", "parents": [ "33d60625faac971ea9d223b64e172f19266dac55" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Fri Jan 30 21:05:29 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jan 30 12:10:34 2026 -0800" }, "message": "[OpenMP] Remove LLVM_ENABLE_PROJECTS\u003dopenmp build mode (#176950)\n\nReapply #152189 and #174963 which were reverted because it broke\npublish-sphinx-docs and publish-doxygen-docs.\n\nThe build mode has been deprecated in #136314 and was supposed to be\nremoved in the LLVM 21 release (#136314).\n\nOpenMP currently supports 4 build modes:\n\n * `cmake \u003cllvm-project\u003e/llvm -DLLVM_ENABLE_PROJECTS\u003dopenmp`\n * `cmake \u003cllvm-project\u003e/llvm -DLLVM_ENABLE_RUNTIMES\u003dopenmp` (bootstrapping build)\n * `cmake \u003cllvm-project\u003e/openmp` (standalone build)\n * `cmake \u003cllvm-project\u003e/runtimes -DLLVM_ENABLE_RUNTIMES\u003dopenmp` (runtimes default/standalone build)\n\nEach build mode increased the maintanance overhead since all build modes\nmust continue working and user confusion when there do not (see #151117,\n#174126, #154117, ...). Let\u0027s finally remove it.\n\nGitOrigin-RevId: c15853ec3a304ff9943dc88424fcb511c3065864\n" }, { "commit": "33d60625faac971ea9d223b64e172f19266dac55", "tree": "2e16d929e10aa6b059f2ac5ba657188a56c3987e", "parents": [ "b0f8ce0b0283ede33df9c0b9c3e9043e22a0a179" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu Jan 29 12:35:42 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jan 29 10:39:24 2026 -0800" }, "message": "[Offload] Provide a cache file for building OpenMP w/ Flang offloading (#178472)\n\nSummary:\nThis build is more annoying, enables everything required for this\nversion of the runtime.\nGitOrigin-RevId: 830236204ec56fe2e268ca3e386e69d8a2a1b114\n" }, { "commit": "b0f8ce0b0283ede33df9c0b9c3e9043e22a0a179", "tree": "88c932a6ecc7de751b67947d39236626679a0810", "parents": [ "142387e3e79154855f1c1e25944b79d8bfb69c23" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Wed Jan 28 14:13:41 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jan 28 05:16:00 2026 -0800" }, "message": "[OpenMP][docs] Revise building manual (#176175)\n\nGo into more detail on the two non-legacy build modes.\n\nI decided use create a dedicated document documention the build process.\n`index.rst` only keeps the \"Getting Started\" part. `README.rst` is\nvastly outdated with the still valid parts integrated into the new\ndocumentation. `SupportAndFAQ.rst` is unstructured and keeps only the\nnon-building parts.\n\nThe the new building document is written in Meltdown, following the\ndecision from LLVM.\n\nGitOrigin-RevId: dd76d3382c5ba4ad050b3e336f72e40ff8b83bcb\n" }, { "commit": "142387e3e79154855f1c1e25944b79d8bfb69c23", "tree": "c22804bcad9394c61a018f345de1d610eae717d3", "parents": [ "7c69203983e49591df5e63e6a24b7147cce99b8e" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue Jan 27 15:55:36 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jan 27 14:00:23 2026 -0800" }, "message": "[OpenMP] Small fixes after refactoring and adding SPIR-V support\n\nGitOrigin-RevId: 5b98c9be179fb838d57e16fc40b2b18b417cf06c\n" }, { "commit": "7c69203983e49591df5e63e6a24b7147cce99b8e", "tree": "e92f3548daddc5c037a56d45e2536644e270a2a1", "parents": [ "af1c3d8a003686de2ed63d6c5a5ea3b6b31fd7c9" ], "author": { "name": "fineg74", "email": "61437305+fineg74@users.noreply.github.com", "time": "Mon Jan 26 16:09:28 2026 -0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jan 26 16:11:03 2026 -0800" }, "message": " [OFFLOAD] Build DeviceRTL with SPIRV backend (#174675)\n\nThis PR adds configuration to build DeviceRTL with SPIRV backend. It is\nprimarily used for level-zero plugin for Intel GPUs\n\n---------\n\nCo-authored-by: Joseph Huber \u003chuberjn@outlook.com\u003e\nGitOrigin-RevId: 8db9774118ccdfc583ed230cbdde1c56567d0643\n" }, { "commit": "af1c3d8a003686de2ed63d6c5a5ea3b6b31fd7c9", "tree": "51861c290b625d8dddf9d00c01e9d244932015d2", "parents": [ "b848f49d1517c7a66f66a28b26c7c79788362b68" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Mon Jan 26 14:20:52 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jan 26 12:25:32 2026 -0800" }, "message": "[OpenMP] Cleanup synchronization primitives (#177710)\n\nSummary:\nThese shouldn\u0027t be so different after we moved away from variants. It\u0027s\nmuch simpler to define this in-line with a single preprocessor\ndefinition. This should be equivalent less a few unnecessary function\ndefinitions with the advantage that SPIR-V now has less work to do.\nGitOrigin-RevId: 55979032401ebdb62085f677770fc6c5793b584e\n" }, { "commit": "b848f49d1517c7a66f66a28b26c7c79788362b68", "tree": "8631c3ed010d5ba2fffdc70f277a9383403d4323", "parents": [ "e93f2a83d262f03aaed35abe2b42b143e22c25c8" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu Jan 22 16:45:07 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jan 22 14:50:17 2026 -0800" }, "message": "[OpenMP] Force initialization of global device environment\n\nSummary:\nThis causes issues when we omit this, which requires a constructor from\nan incompatible address space and bad things happen.\nGitOrigin-RevId: 3219f1a5ce497125025ad8d27131cdea4b6e8595\n" }, { "commit": "e93f2a83d262f03aaed35abe2b42b143e22c25c8", "tree": "4ab66a39be46cc9ed5ab26e61eddae5993b897e7", "parents": [ "6845646cf81a0884063e4f359d6830556f592b53" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu Jan 22 08:02:45 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jan 22 06:06:01 2026 -0800" }, "message": "[Clang] Rename `uinc_wrap` and add normal atomic builtin (#177253)\n\nSummary:\nThe `__scoped_atomic` builtins are supposed to match the standard\nGNU-flavored `__atomic` builtins. We added a scoped builtin without a\ncorresponding standard one before the fork so this should be added in\nthe release candidate. These were originally added in\nhttps://github.com/llvm/llvm-project/pull/168666\n\nAlso, the name `uinc_wrap` does not follow the naming convention. The\nGNU atomics use `fetch_xyz` to indicate that the builtin returns the\nprevious location\u0027s value as part of the RMW operation, which these do.\nThis PR renames it and its uses.\n\nGitOrigin-RevId: d5899ccb6f338e254995760bf11c6acddade5391\n" }, { "commit": "6845646cf81a0884063e4f359d6830556f592b53", "tree": "2c004ad9eb13f71e7a254030fb1d2c011cc49439", "parents": [ "0678ae04244cfaf33919b031c819f863dbcd3d0c" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed Jan 21 11:31:23 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jan 21 09:34:21 2026 -0800" }, "message": "[OpenMP][NFC] Use `uinc` atomic builtins for this operation (#177207)\n\nSummary:\nWe support this now, this is 1-to-1 equivalent and simply prevents us\nfrom needing to do it ourselves.\nGitOrigin-RevId: f534711729b7feb0efc443a5f2139a906a60cf8f\n" }, { "commit": "0678ae04244cfaf33919b031c819f863dbcd3d0c", "tree": "9d14265ca445e24e5e03228e1d87d0984cb84590", "parents": [ "93a5031b1e758f89b7ba67b26e0b3b064859b820" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Tue Jan 20 13:00:54 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jan 20 04:05:44 2026 -0800" }, "message": "Revert \"[OpenMP] Remove LLVM_ENABLE_PROJECTS\u003dopenmp build mode (#174963)\"\n\nThis reverts commit f7b5b670c1cf0fe54e59aa0d1f1a11ee66e442ab.\n\nIt had broken the publish-doxygen-docs buildbot\n\nGitOrigin-RevId: 8acf454ed1b297b239c0fa5901d09597bbbe6da5\n" }, { "commit": "93a5031b1e758f89b7ba67b26e0b3b064859b820", "tree": "6794aac8f9196b3f2fd1ff6694edb9a4adea4eb5", "parents": [ "9cf3533a140c252ecde57f6f6c7f4e09db4bb448" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Tue Jan 20 11:55:43 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jan 20 03:00:34 2026 -0800" }, "message": "[OpenMP] Remove LLVM_ENABLE_PROJECTS\u003dopenmp build mode (#174963)\n\nReapply #152189 which was reverted because it broke publish-sphinx-docs.\n\nThe build mode has been deprecated in #136314. According to the\ndeprecation message, it was supposed to be removed in the LLVM 21\nrelease. Each build mode increased the maintanance overhead when\nfailing, such as in #151117.\n\nGitOrigin-RevId: f7b5b670c1cf0fe54e59aa0d1f1a11ee66e442ab\n" }, { "commit": "9cf3533a140c252ecde57f6f6c7f4e09db4bb448", "tree": "9d14265ca445e24e5e03228e1d87d0984cb84590", "parents": [ "cd3d2cf3f431004a59f0b598061e2e41801da174" ], "author": { "name": "Michael Halkenhäuser", "email": "MichaelGerald.Halkenhauser@amd.com", "time": "Mon Jan 19 20:09:29 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jan 19 11:11:49 2026 -0800" }, "message": "[OpenMP][omptest] Fix CMake target properties (#176802)\n\nFix visibility of target properties:\n * INCLUDE_DIRECTORIES\n * COMPILE_DEFINITIONS\n\nimplicitly pulled in from `default_gtest`.\n\nFixes: https://github.com/llvm/llvm-project/pull/159416\nGitOrigin-RevId: 696b91aeb7e55d5fbc82b2e6a625f2a7ca4a6911\n" }, { "commit": "cd3d2cf3f431004a59f0b598061e2e41801da174", "tree": "a135975afcd227deb0fb13efd1854ccd241b9280", "parents": [ "cfa7f8dc0f8982117e40ce7dee1954186b5f285b" ], "author": { "name": "Michael Halkenhäuser", "email": "MichaelGerald.Halkenhauser@amd.com", "time": "Mon Jan 19 18:40:32 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jan 19 09:45:25 2026 -0800" }, "message": "[OpenMP][omptest] Improve CMake and address review comments (#159416)\n\nAvoid explicit ABI breaking check deactivation\nReplace whole-archive linking with dedicated build of GoogleTest lib\n\nAddresses remaining post-merge review comments of\nhttps://github.com/llvm/llvm-project/pull/154786\n\nGitOrigin-RevId: d5fea7e4f0f2be1f0eae7a3f9e59860632134383\n" }, { "commit": "cfa7f8dc0f8982117e40ce7dee1954186b5f285b", "tree": "9a3ec7584934fe2d4e421f71336d3bdba484ef2c", "parents": [ "af9260a0ed650483afb2b0aa2eecf5494b6c607b" ], "author": { "name": "Aaron Ballman", "email": "aaron@aaronballman.com", "time": "Wed Jan 14 10:17:26 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jan 14 07:20:49 2026 -0800" }, "message": "Nominate Michael Klemm as lead maintainer of OpenMP (#175964)\n\nWe discussed this during the OpenMP sync-up call on 1/14 and he\u0027s still\nactively maintaining this component and would be happy to act as lead\nfor it.\n\nGitOrigin-RevId: bc550d380c76c499d83cc97d23a36a893715602d\n" }, { "commit": "af9260a0ed650483afb2b0aa2eecf5494b6c607b", "tree": "1297cb1a2a5d0ff780f65750c5281dfe909f215f", "parents": [ "e306e5637bef692e7cd44f40fe8c1baae0e78eb2" ], "author": { "name": "Austin Jiang", "email": "austinjiangboyu@gmail.com", "time": "Tue Jan 13 11:52:46 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jan 13 08:57:46 2026 -0800" }, "message": "Fix typos and spelling errors across codebase (#156270)\n\nCorrected various spelling mistakes such as \u0027occurred\u0027, \u0027receiver\u0027,\n\u0027initialized\u0027, \u0027length\u0027, and others in comments, variable names,\nfunction names, and documentation throughout the project. These\nchanges improve code readability and maintain consistency in naming\nand documentation.\n\nCo-authored-by: Louis Dionne \u003cldionne.2@gmail.com\u003e\nGitOrigin-RevId: e6cdfb75ac88719c3bb6f3cab17ec0c8a781a34f\n" }, { "commit": "e306e5637bef692e7cd44f40fe8c1baae0e78eb2", "tree": "db37b52d4459825e94422b3a57d560abdb2efb0f", "parents": [ "32bb09f7230d2bf30659408e32d944c6f362cf22" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Mon Jan 12 16:02:15 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jan 12 07:04:10 2026 -0800" }, "message": "[OpenMP] Add libomp unit test infrastructure (#168063)\n\n(The tests in `TestKmpStr.cpp` are an automatically generated POC to\nmake sure things work.)\n\nGitOrigin-RevId: 26e10cda55d14c1d9ebe6ce4fc36c62eb650ae82\n" }, { "commit": "32bb09f7230d2bf30659408e32d944c6f362cf22", "tree": "f76bd759d966906f1ef77c81f22f889c36e8e348", "parents": [ "6001392d0a79d81f9efa9fbe9d850755ef6ab7f5" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Thu Jan 08 13:10:46 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jan 08 04:16:01 2026 -0800" }, "message": "Revert \"[OpenMP] Remove LLVM_ENABLE_PROJECTS\u003dopenmp build mode (#152189)\"\n\nThis reverts commit 20d0ec849804218d75647aeafbe23f8a02a83b56.\n\nThe publish-sphinx-docs buildbot still uses LLVM_ENABLE_PROJECTS\u003dopenmp.\n\nGitOrigin-RevId: 9ac2d0a9b9ef8893216e8f0b2afb14bcd0a539ff\n" }, { "commit": "6001392d0a79d81f9efa9fbe9d850755ef6ab7f5", "tree": "de2a026f6ce4dc0420738a23abca60797304c6ea", "parents": [ "0de8923afc3055e99c7b7d18bf72a6ed6c2cc2c6" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Thu Jan 08 13:01:07 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Jan 08 04:06:16 2026 -0800" }, "message": "[OpenMP] Remove LLVM_ENABLE_PROJECTS\u003dopenmp build mode (#152189)\n\nThe build mode has been deprecated in #136314. According to the\ndeprecation message, it was supposed to be removed in the LLVM 21\nrelease. Each build mode increased the maintanance overhead when\nfailing, such as in #151117.\n\nLet\u0027s remove it in LLVM 22.\n\nGitOrigin-RevId: 20d0ec849804218d75647aeafbe23f8a02a83b56\n" }, { "commit": "0de8923afc3055e99c7b7d18bf72a6ed6c2cc2c6", "tree": "f76bd759d966906f1ef77c81f22f889c36e8e348", "parents": [ "59f7717744366338c73852512857655414062216" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue Jan 06 08:44:02 2026 -0600" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jan 06 06:45:54 2026 -0800" }, "message": "[OpenMP] Add default stub for virtual methods (#174594)\n\nSummary:\nRecent OpenMP patches have added real support for virtual functions on\nthe device side. However, we don\u0027t provide some of the C++ ABI functions\nthat are emitted when using these. In practice these functions just call\n`std::terminate` so we can just trap here. These are marked weak so they\nwill be overridden by a more correct implementation if not defined and\nwill also not extract on their own from a static library.\nGitOrigin-RevId: 0753286b8e27f995d25b33a4cbd8d07ca0346aa7\n" }, { "commit": "59f7717744366338c73852512857655414062216", "tree": "70c784f8da4b7c600e6e5242869fa6f28bdf534c", "parents": [ "c00e964dc1b01124c3f76a00b6a4142c93494a76" ], "author": { "name": "Joachim", "email": "jenke@itc.rwth-aachen.de", "time": "Thu Dec 18 10:48:30 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Dec 18 01:51:04 2025 -0800" }, "message": "Revert \"[OMPT][Offload][OpenMP] Fixes for OMPT data used by libomptarget\" (#172827)\n\nReverts llvm/llvm-project#156020\n\nWe will need some time for investigating buildbot failures\n\nGitOrigin-RevId: b036d70eeef7b65d889fc9438328e75b7646217e\n" }, { "commit": "c00e964dc1b01124c3f76a00b6a4142c93494a76", "tree": "dc02b2b4c10faa3c26274b6d80a81a38bbecff90", "parents": [ "44f3453aa3ae6603698bcd4321737a66f9a59f6e" ], "author": { "name": "Kaloyan Ignatov", "email": "kaloyan.ignatov@outlook.com", "time": "Thu Dec 18 10:20:14 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Dec 18 01:26:23 2025 -0800" }, "message": "[OMPT][Offload][OpenMP] Fixes for OMPT data used by libomptarget (#156020)\n\nThese commits fix issues regarding storage of tool data within\nlibomptarget. Both libomp and libomptarget have been modified to\naccommodate this. We differentiate between two cases depending on the\ntype of the target region:\n\n- merged target regions (default, without `nowait` clause): behavior\nremains unchanged, tool data is stored in the thread local\nRegionInterface class within libomptarget.\n- deferred target regions (using `nowait` clause): tool data is moved to\n`ompt_task_info_t` struct within libomp, as `RegionInterface` is thread\nlocal and its data is lost whenever another task is scheduled on the\nthread, which happens with deferred target regions.\n\nIn the new implementation, `RegionInterface` receives pointers to\n`ompt_task_info_t` within libomp which are handled transparently within\nlibomptarget. Thus, the problem of tool data getting lost when a thread\nreceives a new task is resolved: `target_data` and `target_task_data`\nremain set.\n\nAnother issue was the value of `task_data` which is supposed to belong\nto the generating task of the region according to the OpenMP standard,\nbut instead had been set to the `task_data` of the target task itself\nuntil now.\n\nTest cases have been added which check both of these fixes.\n\n---------\n\nCo-authored-by: Joachim \u003cjenke@itc.rwth-aachen.de\u003e\nGitOrigin-RevId: 726452720ccf9ba163ebd8d48f660ae6a34a2963\n" }, { "commit": "44f3453aa3ae6603698bcd4321737a66f9a59f6e", "tree": "70c784f8da4b7c600e6e5242869fa6f28bdf534c", "parents": [ "cce521864c6673821e67c251783702566209a34f" ], "author": { "name": "Abhinav Gaba", "email": "abhinav.gaba@intel.com", "time": "Tue Dec 16 15:55:27 2025 -0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:23:08 2025 -0800" }, "message": "[OpenMP][Offload] Add `LIBOMPTARGET_TREAT_ATTACH_AUTO_AS_ALWAYS` to treat `attach(auto)` as `attach(always)`. (#172382)\n\nThis is needed as a way to support older code that was expecting\nunconditional attachment to happen for cases like:\n\n```c\n int *p;\n int x;\n\n #pragma omp targret enter data map(p) // (A)\n #pragma omp target enter data map(x) // (B)\n p \u003d \u0026x;\n\n // By default, this does NOT attach p and x\n #pragma omp target enter data map(p[0:0]) // (C)\n```\n\nWhen the environment variable is set, such maps, where both the pointer\nand the pointee already have corresponding copies on the device, but are\nnot attached to one another, will be attached as-if OpenMP 6.1 TR14\u0027s\n`attach(always)` map-type-modifier was specified on `(C)`.\n\nGitOrigin-RevId: c62cd2877cc25a0d708ad22a70c2a57590449c4d\n" }, { "commit": "cce521864c6673821e67c251783702566209a34f", "tree": "837f5743dbe4b9a2f9e10c50560c328898a93add", "parents": [ "edb8f65cc65157b598771404d23919bbde50d8d7" ], "author": { "name": "Leandro Lupori", "email": "leandro.lupori@linaro.org", "time": "Mon Dec 15 11:25:54 2025 -0300" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:23:03 2025 -0800" }, "message": "[OpenMP] Fix libarcher tests on Ubuntu 22.04 (#170671)\n\nWhen llvm-symbolizer is not found on PATH TSan uses system\u0027s addr2line\ninstead. On Ubuntu 22.04 addr2line can\u0027t handle DWARF v5, which results\nin failures in some libarcher tests.\n\nThis PR adds the directory of the just built LLVM binaries to PATH, to\nmake llvm-symbolizer available to TSan.\n\nThe changes were tested on an AArch64 machine, on which\ntask-taskgroup-unrelated.c was flaky. Moving the test code to a separate\nfunction, executed 10 times, solved the issue.\n\nFixes #170138\n\nGitOrigin-RevId: de776fb60a89f8686fa287ce9b2e8134bf37f436\n" }, { "commit": "edb8f65cc65157b598771404d23919bbde50d8d7", "tree": "d6001e8ef1ce189ce6aa1192bebfd597b67f8317", "parents": [ "7571f4dba600235af945d0fdca19a647ddd14b60" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Tue Dec 09 22:36:44 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:58 2025 -0800" }, "message": "Revert \"[Flang] Move builtin .mod generation into runtimes (Reapply #137828) (#169638)\"\n\nThis reverts commit 7675fc79c802cf7f6a95660f6ee59bf6cb62102f.\n\nRequested in PR:\nhttps://github.com/llvm/llvm-project/pull/169638#issuecomment-3634227707\n\nGitOrigin-RevId: d233e787f0adfa2acd2e6c67aa2c362f9cf47ab4\n" }, { "commit": "7571f4dba600235af945d0fdca19a647ddd14b60", "tree": "2ad2a75f7fa5cee1ac7125b6ffaab83dbd0ffacf", "parents": [ "8a0f6f07a1346f2b3b6db595ff1e03eee025d916" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Tue Dec 09 12:54:26 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:53 2025 -0800" }, "message": "[Flang] Move builtin .mod generation into runtimes (Reapply #137828) (#169638)\n\nReapplication of #137828, changes:\n* Workaround CMAKE_Fortran_PREPROCESS_SOURCE issue for CMake \u003c 2.24: The\nissue is that `try_compile` does not forward manually-defined compiler\nflang variables to the test build environment; instead of just a\nnegative test result, it aborts the configuration step itself. To be\nfair, manually defining these variables is deprecated since at least\nCMake 3.6.\n* Missing flang cmd line flags for CMake \u003c 3.28 `-target\u003d`, `-O2`, `-O3`\n* It is now possible to set FLANG_RT_ENABLED_STATIC\u003dOFF and\nFLANG_RT_ENABLE_SHARED\u003dOFF at the same and is the default for amdgpu and\nnvptx targets. In this mode, only the .mod files are compiled --\nnecessary for module files in\nlib/clang/22/finclude/flang/(nvptx64-nvidia-cuda|amdgpu-amd-amdhsa)/*.mod\nto be available.\n* For compiling omp_lib.mod for nvptx and amdgpu, the module build\nfunctionality must be hoisted out if openmp\u0027s runtime/ directory which\nis only included for host targets. This PR now requires #169909.\n\nMove building the .mod files from openmp/flang to openmp/flang-rt using\na shared mechanism. Motivations to do so are:\n\n1. Most modules are target-dependent and need to be re-compiled for each\ntarget separately, which is something the LLVM_ENABLE_RUNTIMES system\nalready does. Prime example is `iso_c_binding.mod` which encodes the\ntarget\u0027s ABI. Constants such as [`c_long_double` also have different\nvalues](https://github.com/llvm/llvm-project/blob/d748c81218bee39dafb9cc0c00ed7831a3ed44c3/flang-rt/lib/runtime/iso_c_binding.f90#L77-L81).\nMost other modules have `#ifdef`-enclosed code as well. For instance\nthis caused offload targets nvptx64-nvidia-cuda/amdgpu-amd-amdhsa to use\nthe modules files compiled for the host which may contrain uses of the\ntypes REAL(10) or REAL(16) not available for nvptx/amdgpu.\n\n#146876\n#128015\n#129742\n#158790\n\n3. CMake has support for Fortran that we should use. Among other things,\nit automatically determines module dependencies so there is no need to\nhardcode them in the CMakeLists.txt.\n\n4. It allows using Fortran itself to implement Flang-RT. Currently, only\n`iso_fortran_env_impl.f90` emits object files that are needed by Fortran\napplications (#89403). The workaround of #95388 could be reverted (PR\n#169525).\n\nIf using Flang for cross-compilation or target-offloading, flang-rt must\nnow be compiled for each target not only for the library, but also to\nget the target-specific module files. For instance in a bootstrapping\nruntime build, this can be done by adding:\n`-DLLVM_RUNTIME_TARGETS\u003ddefault;nvptx64-nvidia-cuda;amdgpu-amd-amdhsa`.\n\nSome new dependencies come into play:\n* openmp depends on flang-rt for building `lib_omp.mod` and\n`lib_omp_kinds.mod`. Currently, if flang-rt is not found then the\nmodules are not built.\n* check-flang depends on flang-rt: If not found, the majority of tests\nare disabled. If not building in a bootstrpping build, the location of\nthe module files can be pointed to using\n`-DFLANG_INTRINSIC_MODULES_DIR\u003d\u003cpath\u003e`, e.g. in a flang-standalone\nbuild. Alternatively, the test needing any of the intrinsic modules\ncould be marked with `REQUIRES: flangrt-modules`.\n* check-flang depends on openmp: Not a change; tests requiring\n`lib_omp.mod` and `lib_omp_kinds.mod` those are already marked with\n`openmp_runtime`.\n\nAs intrinsic are now specific to the target, their location is moved\nfrom `include/flang` to `\u003cresource-dir\u003e/finclude/flang/\u003ctriple\u003e`. The\nmechnism to compute the location have been moved from flang-rt\n(previously used to compute the location of `libflang_rt.*.a`) to common\nlocations in `cmake/GetToolchainDirs.cmake` and\n`runtimes/CMakeLists.txt` so they can be used by both, openmp and\nflang-rt. Potentially the mechnism could also be shared by other\nlibraries such as compiler-rt.\n\n`finclude` was chosen because `gfortran` uses it as well and avoids\nmisuse such as `#include \u003cflang/iso_c_binding.mod\u003e`. The search location\nis now determined by `ToolChain` in the driver, instead of by the\nfrontend. Another subdirectory `flang` avoids accidental inclusion of\ngfortran-modules which due to compression would result in\nuser-unfriendly errors. Now the driver adds `-fintrinsic-module-path`\nfor that location to the frontend call (Just like gfortran does).\n`-fintrinsic-module-path` had to be fixed for this because ironically it\nwas only added to `searchDirectories`, but not\n`intrinsicModuleDirectories_`. Since the driver determines the location,\ntests invoking `flang -fc1` and `bbc` must also be passed the location\nby llvm-lit. This works like llvm-lit does for finding the include dirs\nfor Clang using `-print-file-name\u003d...`.\n\nGitOrigin-RevId: 7675fc79c802cf7f6a95660f6ee59bf6cb62102f\n" }, { "commit": "8a0f6f07a1346f2b3b6db595ff1e03eee025d916", "tree": "d6001e8ef1ce189ce6aa1192bebfd597b67f8317", "parents": [ "8a0ecdbfd61b72990698d8ce5f1c315a05a124a6" ], "author": { "name": "Kevin Sala Penades", "email": "salapenades1@llnl.gov", "time": "Mon Dec 08 09:37:11 2025 -0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:48 2025 -0800" }, "message": "[Clang][OpenMP] Switch to __kmpc_parallel_60 with strict parameter (#171082)\n\nThis commit switches the `__kmpc_parallel_51` to `__kmpc_parallel_60`,\nand adds the strict boolean for the number of threads.\n\nGitOrigin-RevId: 0e92beb0c0f050e1180c88aa31713f3d6ba9d34e\n" }, { "commit": "8a0ecdbfd61b72990698d8ce5f1c315a05a124a6", "tree": "b325323bf3dc9b0b65f12e99e1e220ba90279f83", "parents": [ "abd57805e4e7b74e6aac33f49c42bc1da50e8af3" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Mon Dec 08 14:15:39 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:44 2025 -0800" }, "message": "[OpenMP] Fix undefined symbol for Darwin builds (#170999)\n\ncf.\nhttps://github.com/llvm/llvm-project/pull/168554#issuecomment-3617253169\n\nGitOrigin-RevId: 33d779dfbf258eac826b2a22b4fece602cce321f\n" }, { "commit": "abd57805e4e7b74e6aac33f49c42bc1da50e8af3", "tree": "dc07c8f27433594bbaebb5179171517337fa3a36", "parents": [ "decdcbfe1c4a0080f3cbdb6ede35e782a2b53824" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Sat Dec 06 12:14:55 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:39 2025 -0800" }, "message": "[openmp][NFCI] Move .mod code out of runtimes subdir (#169909)\n\nExtracted out of #169638. The motivation is that we want to build\nFortran module files for device triples (amdgpu-amd-amdhsa and\nnvptx64-nvidia-cuda) as well, but the runtimes/ directory is only\nincluded for host devices.\n\nThis PR itself should not change anything, including that omp_lib.mod is\nonly built on host devices triple. Some dependencies used for building\nomp_lib.mod are hoisted out of runtimes/ as well. IMHO they all make\nsense to hoist, e.g. LIBOMP_VERSION_MAJOR/LIBOMP_VERSION_MINOR should be\nusable in the entire OpenMP subproject.\n\nGitOrigin-RevId: 830ef4e4e2d0f181ed1cfacbc3daacb2555ff9e4\n" }, { "commit": "decdcbfe1c4a0080f3cbdb6ede35e782a2b53824", "tree": "adf01d0017565a9a09ab8b79da8fb11b6bb4d830", "parents": [ "3da2e1db5ea4179d5c5302fcdb17ab6d7013b826" ], "author": { "name": "Jan André Reuter", "email": "j.reuter@fz-juelich.de", "time": "Wed Dec 03 19:03:34 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:34 2025 -0800" }, "message": "[OpenMP][OMPT] Use global thread id for `codeptr_ra` in `end_critical` (#169826)\n\nWhen a critical construct has finished, it will trigger a\ncritical-released event. If a tool is attached, and the `mutex_released`\ncallback was registered, the tool with receive an event containing the\n`codeptr_ra`, the return address of the callback invocation.\n\nAll the way back in 82e94a593433f36734e2d34898d353a2ecb65b8b, this\n`codeptr_ra` was implemented by calling `__ompt_load_return_address`\nwith a fixed global thread id of `0`. However, this approach results in\na race-condition, and can yield incorrect results to the tool.\n\n`__ompt_load_return_address(0)` points to the current return address of\nthe thread 0 in `__kmp_threads`. This thread may already execute some\nother construct. A tool might therefore receive the return address of\ne.g. some `libomp` internals, or other parts of the user code.\nAdditionally, a call to `__ompt_load_return_address` resets the\n`th.ompt_thread_info.return_address` to `NULL`, therefore also affecting\nthe return address of thread 0. Another dispatched event, e.g.\nparallel-begin might therefore not transfer any `codeptr_ra`.\n\nTo fix this, replace the fixed thread id by the `global_tid`, which is\nstored just before dispatching the `mutex_released` callback.\n\nSigned-off-by: Jan André Reuter \u003cj.reuter@fz-juelich.de\u003e\nGitOrigin-RevId: 94232f9f560f84d2ae7f50b2d1df5bc26b2ce69e\n" }, { "commit": "3da2e1db5ea4179d5c5302fcdb17ab6d7013b826", "tree": "5d055a11e24453c74a966c4a74ca50aedd2e1e31", "parents": [ "a37f53ead831cf52134f551e3975fdbb71ea9f33" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Tue Dec 02 09:54:36 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:29 2025 -0800" }, "message": "Fix Windows OpenMP build (#170142)\n\nfixes Windows build issue in\nhttps://github.com/llvm/llvm-project/pull/168554\n\nGitOrigin-RevId: 96056669493dfd3db653790579b0dbeba0cdaa5f\n" }, { "commit": "a37f53ead831cf52134f551e3975fdbb71ea9f33", "tree": "0ab2a1e54be9048ce9459e60b3a6c14c97625481", "parents": [ "56a257699b2e35c92f334bf394e82fc5ddf52344" ], "author": { "name": "Robert Imschweiler", "email": "robert.imschweiler@amd.com", "time": "Mon Dec 01 14:18:31 2025 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Dec 16 23:22:24 2025 -0800" }, "message": "Reland: [OpenMP] Implement omp_get_uid_from_device() / omp_get_device_from_uid() (#168554)\n\nReland https://github.com/llvm/llvm-project/pull/164392 with Fortran support moved to follow-up PR\n\nGitOrigin-RevId: 8808beeb1a35c8f2ffe228b9e91af5067388f909\n" } ], "next": "56a257699b2e35c92f334bf394e82fc5ddf52344" }